Cell Geometry Dictates Tnfα-Induced Genome Response PNAS PLUS

Cell geometry dictates TNFα-induced genome response PNAS PLUS

Aninda Mitraa,b, Saradha Venkatachalapathya, Prasuna Ratnaa,b, Yejun Wanga, Doorgesh Sharma Jokhuna, and G. V. Shivashankara,b,1

aMechanobiology Institute, National University of Singapore, Singapore 117411; and bInstitute of Molecular Oncology, Italian Foundation for Cancer Research, Milan 20139, Italy

Edited by David A. Weitz, Harvard University, Cambridge, MA, and approved April 10, 2017 (received for review October 31, 2016) Cells in physiology integrate local soluble and mechanical signals to anisotropic (RA, stretched) or circular isotropic (CI, relaxed) state regulate genomic programs. Whereas the individual roles of these for NIH 3T3 fibroblast. We confined cells to these two geometric signals are well studied, the cellular responses to the combined constraints, and stimulated them with TNFα and analyzed the nu- chemical and physical signals are less explored. Here, we investi- clear localization of NFκB (p65), as well as the expression of its gated the cross-talk between cellular geometry and TNFα signaling. downstream target genes. We found that TNFα stimulation led to We stabilized NIH 3T3 fibroblasts into rectangular anisotropic or cir- differential nuclear translocation of NFκB (p65) in cells of different cular isotropic geometries and stimulated them with TNFα and ana- geometries. Depending on cell geometry, TNFα stimulation resulted lyzed nuclear translocation of transcription regulators –NFκB (p65) in actin depolymerization, degradation of IκB (an inhibitor of and MKL and downstream gene-expression patterns. We found that NFκB), and nucleus-to-cytoplasmic shuttling of MKL, a cofactor to TNFα induces geometry-dependent actin depolymerization, which SRF. Activated RNA polymerase II foci were sequestered with MKL enhances IκB degradation, p65 nuclear translocation, nuclear exit of in RA cells, whereas the addition of TNFα resulted in an enrichment MKL, and sequestration of p65 at the RNA-polymerase-II foci. Further, of p65 localization to these sites to regulate gene expression. Im- global transcription profile of cells under matrix-TNFα interplay reveals portantly, whole genome transcriptome maps revealed that cells of a geometry-dependent gene-expression pattern. At a functional level, different geometries show very distinct responses to TNFα.Fur- we find cell geometry affects TNFα-induced cell proliferation. Our re- thermore,weshowthatTNFα-induced cell response depends on cell sults provide compelling evidence that fibroblasts, depending on their geometry. Taken together, our results highlight an important link geometries, elicit distinct cellular responses for the same cytokine. between cytokine signaling with the mechanical state of cells. SCIENCES

cell geometry | TNFα | NFκB | gene expression Results APPLIED PHYSICAL Cell-Geometric Constraints Modulate TNFα-Induced Nuclear Translocation o determine their behavior, cells living in multicellular systems of p65. This study aimed to understand the role of cell-geometric α integrate both local soluble and mechanical signals. In recent constraints on signaling events induced by the cytokine TNF .We T α years, evidence has revealed the role of the extracellular mechanical compared the response of NIH 3T3 fibroblasts to TNF , when environment in regulating nuclear signaling, as well as driving gene- confined to two distinct mechanical states, achieved using fibro- –– expression programs that regulate cellular homeostasis (1, 2). In nectin micropatterns. The two chosen fibronectin micropatterns μ 2 μ 2 response to mechanical stimuli induced by the extracellular matrix, rectangles (area: 1,800 m , aspect ratio 1:5) and circles 500 m specific intracellular signaling pathways can be triggered, leading to (Fig. 1A and SI Appendix,Fig.S1A, a), render contrasting features ∼ μ 2 nuclear shuttling of transcription regulators like YAP/TAZ, MKL, to the NIH 3T3 fibroblasts (with average spread area 1,400 m and thereby triggering the expression of specific genes (3–5). Me- in unpatterned conditions), in terms of their actomyosin contrac- chanical properties of the extracellular matrix can thereby regulate tility, spread area, and polarization of F-actin filaments. The cell behaviors, including stem-cell lineage specification, embryonic rectangles give the fibroblasts an anisotropic and stretched shape, development, and also tumor progression or repression (6–9). with a well-spread morphology having a high degree of actin po- Recently, we showed that cell-geometric constraints result not only lymerization and myosin contractility, whereas the circles render in cytoskeletal remodeling (10), but also modular changes in gene them with an isotropic, or symmetrical shape, with a relaxed, less expression. For example, rectangular cells turned on SRF target genes, whereas circular cells turned on NFκB target genes (3). Significance Over the last 30 y, a large body of work has mapped and characterized signaling pathways induced by major soluble signals like Cells experience distinct forces within the tissue microenviron- tumornecrosisfactorα (TNFα) and transforming growth factor β ment, and their geometry-dependent differential responses to (TGFβ). TNFα is a proinflammatory cytokine that is released various cytokines are important in the maintenance of their cel- primarily by activated macrophages, as well as lymphoid cells, mast lular homeostasis. Using micropatterned substrates to alter cell cells, endothelial cells, fibroblasts, and neuronal tissue (11). TNFα geometry, we show that TNFα stimulation results in differential functions through the canonical NF-κB pathway, stress kinases, nuclear localization of the downstream transcription factors and and in some cells, the apoptotic caspase pathway (11). TNFα- modular changes in gene-expression patterns. Our results high- dependent activation has been shown to result in actin remodeling light the importance of the intrinsic geometric properties of a cell (12, 13) and NF-κB-dependent transcription of anti-apoptotic and in determining its response to a biochemical signal. Alterations to proinflammatory genes (14–16). Well-regulated responses to this cellular homeostasis may result in physiological abnormalities TNFα and NF-κB activation are important to normal physiology at single-cell level, leading to diseases like fibrosis and cancer. (17–19). However, the effect of integrating both mechanical and Author contributions: A.M., S.V., P.R., Y.W., D.S.J., and G.V.S. designed research; A.M., P.R., biochemical inputs on nuclear signaling and gene expression and Y.W. performed research; S.V. and D.S.J. contributed new reagents/analytic tools; A.M., programs is not well understood. This is particularly important, S.V., P.R., Y.W., D.S.J., and G.V.S. analyzed data; and A.M., S.V., and G.V.S. wrote the paper. because changes in mechanical inputs regulate cytoskeletal and The authors declare no conflict of interest. nuclear architecture (4, 6, 7, 9, 20), thereby giving a spatial di- This article is a PNAS Direct Submission. mension to gene regulation (21) induced by soluble factors. Freely available online through the PNAS open access option. In this paper we study at single-cell resolution the cross-talk be- 1To whom correspondence should be addressed. Email: [email protected]. α tween cellular geometricconstraintsandTNF-mediated nuclear This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10. signaling. We used micropatterned substrates to define a rectangular 1073/pnas.1618007114/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1618007114 PNAS Early Edition | 1of10 Downloaded by guest on September 27, 2021

A Reactangular Anisotropic (RA) Circular Isotropic (CI) F-actin

BC Utr Utr TNF TNF

p65 4 ***

3 * F-actin

2 p65 N/T p65

DAPI 1 (Normalized w.r.t. RA cells) RA (Normalized w.r.t.

Merge Utr TNF

DEF

40 5 3.5 1.E-01 *** 4

max (min) max 20 t 3.0

4 5.E-02 Slope

0 0.E+00 2.5

3 3 p65 N/T p65 p65 N/T p65 2.0 2 2

1.5 Fold change (p65 N/T) (p65 change Fold (Normalized w.r.t. RA cells) RA (Normalized w.r.t. (Normalized w.r.t. RA cells) RA (Normalized w.r.t.

max max

t t 1 1.0 0204060 0 5 10 15 20 25 TNF (mins) TNF (mins) TNF/Utr GJHI

2.5 ** ** 2.0 1 1.5 2.0

1.5 (cytoplasmic) 1.5 (cytoplasmic) 1

0.5 (cytoplasmic)

1.0 1.0 (cytoplasmic)

B B I

κ 0.5

B B I κ 0.5

0.5

B B phospho-I κ

B B phospho-I κ 0.0 0 0 010200510 TNF (mins) TNF + MG132 (mins)

Fig. 1. Cell-geometric constraints modulate TNFα-induced nuclear translocation of p65. (A) Three-dimensional projections of confocal sections of NIH 3T3 cells plated on fibronectin micropatterns. Rectangles: 1,800 μm2 (AR 1:5); circles 500 μm2, fixed and stained with phalloidin (red) and DAPI (blue). (Scale bar, 10 μm.) (B) NIH 3T3 cells on fibronectin micropatterns. Rectangles: 1,800 μm2 (RA); circles: 500 μm2 (CI); untreated (Utr) or treated with TNFα (20 ng/mL, 30 min) (TNF) stained with p65 (green), phalloidin (red), and DAPI (blue). (Scale bar, 10 μm.) (C) Box plots of fluorescence intensities of p65. N/T of RA and CI cells treated and stained as in B and normalized with respect to untreated RA cells. Data are presented as mean ± SD; box plot represent 25–75 percentile; the small square within each box indicates mean; line indicates median ***P < 0.0001; *P = 0.0058; Student’s t test; n = 85–96 cells in each conditions. (D) Quantifications of average fluorescence intensities of p65. N/T of RA and CI cells, fixed at different time points after stimulation with TNFα, and stained for p65; n = 25–30 cells per condition per time point. (E) Scattered plot showing p65 N/T for the first 25 min of TNFα stimulation. The straight lines indicate linear-fitted curves of the average p65 N/T. (Inset) Slope of the curves for the cells of the two geometries. (F) Box plots of average fold change (TNFα treated: untreated) of p65 N/T of RA and CI cells. ***P < 0.0001; Student’s t test. Box plots of fluorescence intensities (cytoplasmic fraction; normalized with average intensity of RA cells) of (G)IκBinRAandCIcells;n = 13 cells per condition; Student’s t test; **P = 0.0031 and (H)phospho-IκB (S32, S36) in RA and CI cells; n = 12–14 cells per condition; Student’s t test; **P = 0.0091. Data are presented as mean ± SD; box plots represent 25–75 percentiles; the small square within each box indicates mean; line indicates median. (I) Scattered line plots of the average fluorescence intensities of IκB (cytoplasmic fraction; normalized with average intensity of RA cells) of RA and CI cells fixed at different time points after stimulation with TNFα and stained for IκB. Data are presented as mean ± SD; n = 13–15 cells per condition per time point. (J) Scattered line plots of the average fluorescence intensities of phospho-IκB (cytoplasmic fraction; normalized with average intensity of RA cells) of RA and CI cells fixed at different time points after stimulation with TNFα in the presence of proteasome inhibitor: MG132 and stained for phospho-IκB antibody that recognizes phospho-Ser32 and phospho-Ser36. Data are presented as mean ± SD; n = 12–16 cells per condition per time point.

2of10 | www.pnas.org/cgi/doi/10.1073/pnas.1618007114 Mitra et al. Downloaded by guest on September 27, 2021 spread morphology having a low degree of actin polymerization observed. These results collectively suggest that TNFα-mediated PNAS PLUS and contractility (Fig. 1A and SI Appendix,Fig.S1A, b–g)asin- differential processing of IκB and subsequent p65 nuclear locali- dicated by the cytoplasmic levels of F-actin and phospho-myosin- zation is dependent on cell geometry. Intriguingly, we find that CI light-chain (pMLC). We aimed to determine how mechanical cells have higher efficiency of nuclear shuttling of p65 for similar constraints, by themselves or in conjunction with TNFα signaling, amounts of degraded IκB, as observed by plotting the changes in can regulate nuclear shuttling of transcription regulators. To study p65 nuclear fraction as a unit of IκBchange(Δp65/ΔIκB) across the response generated by TNFα signaling, we assessed the extent time points of TNFα stimulation (SI Appendix,Fig.S1C, e). of nuclear translocation of its effector, NFκB (p65 subunit). In Because differential biochemical processing can occur as a result NIH 3T3 cells grown in unstimulated conditions, on unpatterned of differences in the mobility of the molecules in question, we fibronectin p65 localization was observed to be cytoplasmic. looked at IκB mobility. The mobility of IκB was studied using However, within 30 min of TNFα (20 ng/mL) stimulation the fluorescence recovery after photobleaching (FRAP) experiments, p65 was seen to translocate to nucleus (SI Appendix,Fig.S1B, a with mCherry-IκB expressed in NIH 3T3 cells under the two and b). To explore the effect of cell geometry on TNFα-induced geometric constraints. Regions of interest of 2.5-μmdiameterin cellular response, we plated NIH 3T3 cells on either RA or CI the cytoplasmic region of cells on the two shapes were photo- fibronectin micropatterns (Materials and Methods), which were left bleached and the recovery of the intensities of mCherry was untreated (Utr) or treated with TNFα (20 ng/mL, 30 min) (TNF) tracked over time. mCherry-IκB mobility was higher in circular before being fixed and immuno-stained with p65 antibody and cells (having a lower t1/2 recovery) than the RA cells (having a phallolidin (to stain F-actin). Images of single cells, acquired using higher t1/2 recovery) (SI Appendix,Fig.S1D). Thus, the mobility of confocal microscopy (Fig. 1B and SI Appendix,Fig.S1B, c and d), IκB in the cytoplasm is differentially modulated by the mechanical were analyzed to determine a nucleus-to-total ratio (N/T) of p65 state of the cells, and this may affect the biochemical processing (SI Appendix,Fig.S1B, h), and plotted after normalization with induced by TNFα signaling. Whereas IκB degradation is highly respect to untreated cells on RA patterns (Fig. 1C). correlative to p65 translocation, there may be other layers of reg- Cells on CI patterns have slightly higher levels of nuclear p65 ulation that modulate the nuclear entry of p65. To understand this (N/T) compared with those on RA patterns (SI Appendix, Fig. S1 cell-geometry-dependent differential processing of TNFα signal, B, e and f). This indicates that the mechanical state of the cell alone we explored the role of the actin cytoskeleton which is known to be can affect the nuclear levels of the transcription factor p65. Upon different in the RA and the CI cells. α α stimulation with TNF , CI cells showed a higher extent of TNF - SCIENCES induced p65 nuclear translocation compared with RA cells (Fig. 1 TNFα Stimulation Results in Actin Depolymerization in RA Cells to B and C and SI Appendix,Fig.S1B, h and i). A time-course study Modulate Nuclear-Cytoplasmic Compartmentalization of p65 and MKL. APPLIED PHYSICAL of TNFα stimulation on p65 nuclear translocation revealed that the Fibroblast cells of the two geometries differ in terms of their ac- CI cells show higher p65 (N/T) at various time points from 0 to tomyosin contractility. In RA cells, the intensity levels of cytoplas- 25mincomparedwithRAcells(Fig.1D and E and SI Appendix, mic F-actin were higher than in CI cells, as RA cells possess robust Fig. S1 B, k and l). The slope of the linear-fitted curve for the first F-actin fibers (Fig. 2A and SI Appendix,Fig.S2A). Furthermore, the 25 min is higher for CI cells, indicating faster nuclear translocation amount of pMLC was higher in RA cells, which indicates higher of p65 compared with RA cells (Fig. 1E). Nuclear translocation of myosin activity compared with that in CI cells. This in turn suggests NFκB is known to be negatively regulated by its inhibitor, IκB, a higher Rho activity in RA cells (SI Appendix,Fig.S2B, e). which is one of the early target genes to be transcribed. We observe The differences in levels of cytoplasmic F-actin and pMLC in that the nuclear exit of p65 occurs earlier and faster in CI cells polarized and CI cells prompted us to look at the Rho GTPase (with peak time tmax = 25 min) in comparison with RA cells (tmax = signaling pathway. In cells, Rho GTPase-mediated signaling is 35 min) (Fig. 1D). Moreover, the average fold change of nuclear known to regulate actin polymerization and myosin contractility by p65 before and after 30 min of TNFα stimulation is also signifi- modulating ROCK activity and myosin phosphorylation (23). cantly higher in CI cells compared with the RA cells (Fig. 1F). F-actin severing can occur through the activity of the cofilin/ADF These observations suggest that the signaling events leading to family of proteins, which are regulated by ROCK and LIM kinases. TNFα-induced nuclear translocation of p65 is faster in CI cells. Specifically, LIM kinase-2 is known to phosphorylate the Serine-3 Nuclear translocation of the NFκB dimer occurs due to the residue of cofilin, and thereby regulate its activity by deactivating proteasome-mediated degradation of IκB. This follows IκB phos- the protein (24–26). Conversely, dephosphorylation of S3 leads to phorylation by IκB kinase, which is induced by TNFα (22). The its activation. The RA cells possess higher levels of cytoplasmic cytoplasmic level of IκB, as well as its phosphorylation levels phospho-LIM kinase-2 (SI Appendix,Fig.S2B, f) and phospho- (phospho-Serine-32 and -36), were analyzed in cells grown on the cofilin (SI Appendix, Fig. S2 B, g) (compared with CI cells. This two shapes. In unstimulated conditions, RA cells have higher cy- indicates that the activity of LIM kinase-2 is higher and the activity toplasmic levels of IκB(Fig.1G and SI Appendix,Fig.S1C, a and of cofilin is lower in RA cells. These factors can contribute toward b), and lower levels of cytoplasmic phospho-IκB, compared with CI higher F-actin in RA cells compared with CI cells. cells (Fig. 1H and SI Appendix,Fig.S1C, c and d). Upon stimu- We next explored how these parameters change upon stimulation lation with TNFα, there is rapid degradation of IκBinRAcells. with TNFα.WeobservedarapiddecreaseinF-actinintensityinRA TheamountofIκB degraded in the first 10 min of TNFα stimu- cellsovertime,afterTNFα stimulation (Fig. 2B and SI Appendix,Fig. lation is much higher in RA cells than in CI cells (Fig. 1I and SI S2 A, c). In CI cells, however, the F-actin intensity was low to begin Appendix,Fig.S1C, a and b). with and we observed a decrease in F-actin intensity upon TNFα Because phosphorylation and polyubiquitination of IκBprecede stimulation, but to a much lower extent than was observed in RA cells. its degradation, the phosphorylation levels of IκB were measured Our observation that TNFα stimulation causes changes in the across a time course of TNFα stimulation. Cells on the two shapes levels of F-actin directed us to estimate the fate of Rho activity and were stimulated with TNFα (20 ng/mL) in the presence of the downstream signaling components before and after TNFα stimu- proteasome inhibitor MG132, to block IκB degradation, to enable lation. RA cells possess higher levels of active-Rho in the cyto- an accurate estimation of phosphorylation levels. The cells were plasm compared with CI cells. TNFα stimulation for 30 min leads then fixed at different time points and immunostained for phos- to a decrease in active-Rho levels predominantly in RA cells with a pho-IκB (phospho-Serine-32 and -36). Upon stimulation with marginal decrease in CI cells (Fig.2C and SI Appendix,Fig.S2B, TNFα, a rapid increase in phosphorylation of IκB was observed in a). The TNFα-induced Rho inhibition is supported by the observed RA cells, compared with CI cells (Fig. 1J and SI Appendix,Fig.S1 time-dependent decrease in phospho-MLC levels upon TNFα C, c and d) where marginal increase in phosphorylation of IκBwas stimulation in RA cells, which showed little change in CI cells (SI

Mitra et al. PNAS Early Edition | 3of10 Downloaded by guest on September 27, 2021

A Utr TNF Utr BC

yt )n isn ***

nitca-F 1.5

oitc

etni xa etni

nitc A

ejorp 1

-F R

1.0

M( evitc

0.5 a 0.5

Apical 0 0.0 0 5 10 15 20 Utr TNF Utr TNF TNF (mins) D E 1

Middle 1

2KM

ilif

ILp

C- p 0.5 0.5

Basal

0 0 01020 0102030 TNF (mins) TNF (mins) FG H I J KL 4 *** *** ***

*** 2.0 GFP 5

*** ) 2

1.5 r

GFP-RhoV14 o

4 )cims

s g

NS tc n

1.2 e

n )

) 3

)

af tn af

e eg

eg 1 h

o 4 hc

Utr No. of

n-n

(fold change)

1.5 a

actin-nodes

ahc d ahc

dl e

1.0 l

hc hc hc hc

tc 0

of otyc( otyc(

m 2

(

c )T/N

l 0.8

l i

2 f

of(

/N(

( ( (

1.0 .

B (cytoplasmic) (cytoplasmic) B

κ 0.5

(

I I

TNF

1 B

N 2

κ I

on actin-node s (fold change) p) p 0.4 0.5 0.0 0 1 0 01020 Utr CytoD Utr CytoD Utr TNF Utr TNF 0510 TNF (mins) GFP GFP-RhoV14 TNF (mins) M N O P *** Utr Utr TNF TNF nodes 6 0.6 1.2 B on 8 4 *** )

cytosol κ

e g

p-I n

actin actin-nodes 1 MKL

2 a hc 6 L T/N

IPA

0.4 o

f K

( T/N ( 0.8 M

4 D

e 0.6

g M

2 reM 0.2 phospho-IkB enrichment factor factor enrichment phospho-IkB 0.4 0 Utr TNF Utr TNF 0102030 0246810 TNF (mins) TNF (mins)

Fig. 2. TNFα stimulation results in actin depolymerization in RA cells to modulate compartmentalization of p65 and MKL. (A) Maximum-intensity projections of confocal sections (0.5 μm) of RA and CI cells before and after 30-min TNFα stimulation, stained for F-actin (phalloidin). (Scale bar, 10 μm.) Lower panel shows the basal, middle, and apical sections of these cells. (Scale bar, 10 μm.) (B) Scattered line plots of the average fluorescence intensities of the cytoplasmic fractions of F-actin (normalized with average intensity of untreated RA cells) in RA and CI cells fixed at different time points after stimulation with TNFα and stained for phalloidin. Data are presented as mean ± SD; n = 12–14 cells per condition per time point. (C) Box plots of fluorescence intensities of active RhoA (cytoplasmic fraction; normalized with average intensity in RA cells) of RA and CI cells fixed and stained for active-RhoA. Data are presented as mean ± SD; box plot represent 25–75 percentile; the small square within each box indicates mean; line indicates median ***P < 0.0001; n = 15–21 cells per condition. Scattered line plots of the average fluorescence intensities of the cytoplasmic fractions of (D) phospho-LIM-kinase2 and (E) phospho-Cofilin-S3 (normalized with average intensity of untreated RA cells) in RA and CI cells fixed at different time points after stimulation with TNFα. Data are presented as mean ± SD; n = 13–20 cells per condition per time point. Box plots of fluorescence intensities of (F)IκB (cytoplasmic fraction); n = 15–21 cells per condition. (G) p65, N/T of RA cells with or without Cytochalasin D treatment (200 nM, 30 min); n = 30 cells per condition. (H) Scattered line plots of fluorescence intensities of p65, N/T of RA cells expressing either GFP (red) or GFP-RhoV14 (orange) fixed at different time points after stimulation with TNFα and stained for p65. (I) Box plots of fluorescence intensities of IκB(cytoplasmic fraction; normalized with average intensity of untreated RA cells) in untreated and TNFα-treated cells expressing either GFP (red) or GFP-RhoV14 (black). n = 10– 12 cells per condition ***P < 0.0001; Student’s t test. N.S., not significant. (J) Representative images showing segmentation (Materials and Methods) of actin nodes (yellow) in untreated and TNFα-treated RA cells stained for F-actin (gray). (K) Quantification of number of actin nodes in RA cells (fold change of untreated cells) across time points of TNFα stimulation. (Inset) Fold difference of the number of actin nodes in untreated cells of the two geometries. (L)IκBenrichmentfactoron the actin nodes in untreated cells of the two geometries. (M)Phospho-IκB enrichment factor on the actin nodes, cytosol, and the total actin of RA cells (fold change of untreated cells) across time points of TNFα stimulation. (Inset)Folddifferenceofphospho-IκB enrichment factor in untreated cells of the two geometries. (N) Representative images of RA and CI either untreated (Utr) or stimulated with TNFα for 30 min (TNF) stained with MKL (green) and DAPI (blue). (Scale bar, 10 μm.) (O) Box plots of the fluorescence intensities of MKL; N/T of cells, treated and stained as in N.***P < 0.001; n = 12–18 cells per condition. (P) Scattered line plots of the average fluorescence intensities of MKL; N/T of RA and CI cells (fold changes of untreated RA cells) at different time points after stimulation with TNFα; n = 12–18 cells per condition per time point.

Appendix, Fig. S2 B, b and e). We also found that upon TNFα To understand the implication of TNFα-induced actin de- stimulation, phospho-LIM kinase-2 levels (Fig. 2D and SI Appen- polymerization in the canonical NFκB pathway, we explored if actin dix,Fig.S2B, c and f) and phospho-cofilin levels (Fig. 2E and SI depolymerization was required for IκB processing and NFκBnu- Appendix, Fig. S2 B, d and g) decrease rapidly in RA cells. In clear translocation. We found that TNFα-induced p65 nuclear contrast, there is no significant change in phospho-LIM kinase-2 translocation was significantly enhanced in RA cells when they were and phospho-cofilin levels in CI cells (Fig. 2 D and E and SI Ap- pretreated for 15 min with drugs that inhibit F-actin polymerization pendix,Fig.S2B, c, d, f,andg). This indicated that TNFα-induced and myosin activity (CytochalasinD, LatrunculinA, or Blebbistatin) actin depolymerization is possibly due to inhibition of Rho leading (SI Appendix,Fig.S2C). RA cells when treated with only Cytocha- to an increase in cofilin activity. lasinD showed a decrease in the levels of cytoplasmic IκB(Fig.2F

4of10 | www.pnas.org/cgi/doi/10.1073/pnas.1618007114 Mitra et al. Downloaded by guest on September 27, 2021 and SI Appendix,Fig.S2D, a) and an increase in nuclear trans- leads to MKL exit from the nucleus, whereas there was marginal PNAS PLUS location of NFκB(Fig.2G and SI Appendix,Fig.S2D, b)indicating change in nuclear localization of MKL in CI cells (Fig. 2 N and O). that actin depolymerization alone was sufficient to drive IκBdeg- A time-course study on the effect of TNFα stimulation on the radation and NFκB translocation even without TNFα stimulation. nuclear levels of MKL revealed that in RA cells MKL rapidly To determine whether the TNFα-induced depolymerization of translocates to the cytoplasm (Fig. 2P), which supports our obser- F-actin via inhibition of Rho is a requirement for TNFα-mediated vation that TNFα induces actin depolymerization in RA cells. canonical-signaling events, we compared p65 nuclear translocation, Thus, we show that TNFα, a cytokine that is known to cause nu- and degradation of IκB after TNFα stimulation, in rectangular cells clear translocation of its effector, the transcription factor NFκB expressing either constitutively active Rho (GFP-RhoV14) or GFP (p65) by a canonical pathway, can also lead to the nuclear exit of (control) (SI Appendix,Fig.S2D, c). another transcription cofactor, MKL. As a result, the TNFα-cell- RA cells expressing GFP-RhoV14 showed slower and compro- geometry interplay creates nuclear-cytoplasmic compartmentaliza- mised p65 nuclear translocation compared with those expressing tion of the two transcription regulators −NFκB (p65) and MKL. GFP (Fig. 2H). Supporting this observation, we found that in RA cells expressing GFP-RhoV14, the degradation of IκB was inhibited TNFα Cell-Geometry Interplay Rewires Transcription Machinery (Fig. 2I). This indicates that TNFα-induced Rho inhibition is not Leading to Differential Expression of p65 and MKL Target Genes. just an additional consequence, but is required for the canonical To check whether these transcription factors/cofactors were nuclear shuttling of p65. Thus, we demonstrate that F-actin de- recruited to active RNA polymerase II pockets (29), we colabeled polymerization is associated with TNFα-mediated nuclear trans- active 5S RNA polymerase II (phospho CTD Serine5) with either location of p65, whereas inhibition of F-actin depolymerization/ MKL or p65. Three-dimensional confocal images of active 5S Rho inhibition inhibits the same, and thereby connects the actin RNA polymerase II and the transcription factors were thresholded remodeling to canonical TNFα signaling. to remove background noise (SI Appendix,Fig.S3A, a and b). The We next quantified the changes in F-actin structures as a con- colocalization fraction was then measured to indicate the coloc- sequence of cell geometry and TNFα. Low doses of actin de- alization of MKL or p65 with active 5S RNA polymerase II (depolymerization agents (Latrunculin A) have been shown to induce scribed in Materials and Methods). This was not sensitive to the actin nodes (27).We segmented the F-actin into actin nodes and degree of image thresholding within a proper range (SI Appendix, filaments (Fig. 2J and SI Appendix,Fig.S2E and F) followed by Fig. S3 A, f and g). We found that in RA cells, the colocalization

quantification of the number of actin nodes. We found that in fraction between MKL and active 5S RNA polymerase II was SCIENCES unconstrained NIH 3T3 cells, F-actin depolymerization induced by ∼50%. This dropped to ∼1% after deletion of the SRF binding TNFα treatment led to appearance of actin nodes similar to when domain of MKL (SI Appendix, Fig. S3 A, c and e). This indicated APPLIED PHYSICAL treated with actin-depolymerization drugs like CytochalasinD (SI that the colocalization between MKL and active 5S RNA poly- Appendix,Fig.S2E). CI cells were found to have a higher amount merase II required MKL to be bound to SRF. This is reminiscent of such nodes than RA cells in untreated condition (Fig. 2 A and of the role of MKL as a cofactor of SRF for gene activation. K). We observed an increase in the number of nodes upon TNFα In CI cells, the colocalization of MKL and 5S RNA polymerase stimulation in RA cells (Fig. 2K). II was dramatically low. This was expected given the cytoplasmic We also observed that the TNFα-generated actin nodes were localization of MKL in these cells. TNFα induction in RA cells also enriched with IκB (Fig. 2L and SI Appendix, Fig. S2G). We hy- reduced the number of MKL/5S RNA polymerase II clusters. This pothesized that these actin nodes could serve as scaffolds for IκB was slightly rescued upon treating cells with Leptomycin B (LMB), binding and processing of signal. Using an FRAP assay, we found a nuclear export inhibitor (Fig. 3 A and B). This is consistent with that IκB mobility was slower on F-actin nodes (SI Appendix,Fig. the observed blockage of TNFα-induced nuclear-to-cytoplasmic S2 H, a–c) than it was in the cytosol. However, IκB mobility on shuttling of MKL in RA cells upon LMB treatment (SI Appen- F-actin filaments did not differ significantly from that in the cytosol dix,Fig.S3B). (SI Appendix, Fig. S2 H, d–f). We found that upon TNFα stim- Consistent with the nuclear enrichment of p65 in CI cells, a ulation, phosphorylation of IκB primarily increased on the actin significantly higher level of p65 and 5S RNA polymerase II clusters nodes with time, whereas in the cytosol IκB phosphorylation was observed in these cells. In RA cells, TNFα stimulation in- increased marginally with time (Fig. 2M). In unstimulated con- creased the colocalization of p65 with 5S RNA polymerase II to a ditions, CI cells have higher phospho-IκB enrichment on actin level that was comparable to that in CI cells. LMB treatment on the nodes than that in RA cells (Fig. 2M, Inset). This suggests that TNFα-stimulated RA cells further enhanced the p65 and 5S RNA the TNFα-generated actin nodes may serve as scaffolds for IκB polymerase II colocalization (Fig. 3 A and C). This is consistent binding, and biochemical events like phosphorylation related to with the observation that TNFα-induced nuclear translocation of IκB processing. This could explain why IκB degradation and p65 in RA cells was further enhanced upon LMB treatment (SI nuclear shuttling of p65 was affected by the enhancement or Appendix,Fig.S3B, a and c). inhibition of F-actin depolymerization. Because TNFα treatment promoted a cell-geometry-dependent The observation that TNFα-induced actin depolymerization differential nuclear localization of the two-transcription regulators could regulate nuclear translocation of NFκB (p65) prompted us to p65 and MKL, we explored the functional implication of this explore the fate of other actin-dependent transcription regulators. phenomenon by assessing the expression of a target gene. In fi- Nuclear shuttling of the SRF transcription cofactor MKL is known broblasts, α-Smooth-Muscle-Actin (αSMA) is a known candidate to be regulated by actin polymerization. Earlier work in the liter- gene which is regulated by MKL, and whose expression level is ature has elucidated the nuclear fraction of MKL to be highly physiologically important, especially during transdifferentiation of sensitive to F-actin/G-actin ratio (28). MKL binds to G-actin, and fibroblasts to myofibroblasts (30). Real-time quantitative PCR hence localizes at the cytoplasm when there is less polymerized (qPCR) studies using mRNA from the RA and CI fibroblast cells, actin. When actin polymerization increases, MKL uncouples from either untreated or treated with TNFα (20 ng/mL, 30 min), showed G-actin and translocates to the nucleus. We therefore monitored that RA cells have higher expression levels of αSMA compared MKL (N/T) in RA and CI cells, over a time course of TNFα with the CI cells. TNFα stimulation led to a significant decrease in stimulation. RA cells showed a high amount of MKL in the nu- αSMA expression levels in RA cells alone, with CI cells not cleus, which is consistent with the presence of robust F-actin fibers responding to TNFα stimulation to a significant degree (SI Ap- in these cells (Fig. 2N). CI cells, on the other hand, have very low pendix,Fig.S3C, a). LMB was used to trap both p65 and MKL in levels of MKL in the nucleus, consistent with the low amount of the nucleus upon TNFα stimulation. qPCR studies on the ex- F-actin observed in the cytoplasm. In RA cells, TNFα stimulation pression levels of αSMA in TNFα-treated RA cells, with or without

Mitra et al. PNAS Early Edition | 5of10 Downloaded by guest on September 27, 2021 ABMKL, phospho-RNA pol II, nucleus p65, phospho-RNA pol II, nucleus C

3.0 ** 2.0 *** *** 2.5 *** NS 1.5 2.0

1.5 1.0

1.0 0.5 0.5 TNF 0.0 0.0 normalised colocalization factor normalised colocalization factor p65 & phospho-RNA polymerase II p65 & phospho-RNA MKL & phospho-RNA polymerase II & phospho-RNA MKL Utr Utr TNF TNF Utr Utr TNF TNF TNF + LMB + LMB + LMB

DE Utr TNF Utr Row Z-score (Summed Z-score) TNF

Relative gene expression Utr TNF Utr TNF NFκB target genes FG Utr TNF Utr Row Z-score (Summed Z-score) TNF Utr TNF Utr TNF MKL-dependent SRF target genes Relative gene expression

Fig. 3. TNFα–cell-geometry interplay rewires transcription machinery leading to differential expression of p65 and MKL target genes. (A) Representative images of the nucleus of RA or CI cells stained for active-RNA-polymerase-II (phospho S5) (pol2) along with either MKL or p65. (Scale bar, 5 μm.) Box plots quantifying the normalized colocalization factor of (B) MKL or (C) p65 with active 5S RNA polymerase-II. Data are presented as mean ± SD with 20 < n < 30. ***P < 0.001, **P < 0.01; N.S., not significant; Two-sample Student’s t test. Heat maps of row Z scores indicating the relative expression of all (D)NFκB target genes and (F) MKL-dependent SRF target-genes in RA and CI cells. Data represented as median of three biological replicates. The summed Z score indicating the total gene expression of all (E)NFκB target genes and (G) MKL-dependent SRF target genes in RA and CI cells. Data represented from three biological replicates. The list of genes in D and F has been tabulated in SI Appendix, Tables S2 and S3.

LMB, showed that the decrease in TNFα-induced αSMA expres- cates that the cell geometry plays a role in interpreting the cellular sion was rescued in the presence of LMB (SI Appendix, Fig. S3 C, response to TNFα stimulation. b). We next assessed the functional implications of the differential Consistently, the MKL-dependent SRF target genes are nuclear shuttling of the transcription regulators by assessing the expressed at relatively higher levels in RA cells before TNF stim- total transcription in cells held in the two geometric states, and ulation compared with CI cells and, upon TNF stimulation, the upon integration of TNFα using microarrays (see SI Appendix, expression of these genes further reduces in both the geometries Materials and Methods and controls in SI Appendix,Fig.S3D). (Fig. 3 F and G). Similar to NFκB target genes the overall trend of Our observations of cell geometry and TNF-mediated com- the expression pattern of the MKL-dependent SRF target genes partmentalization of transcription factors prompted us to explore correlated with the accumulation of the transcription factor in the the expression profiles of the target genes of these transcription nucleus (SI Appendix,Fig.S3K, b). In addition, as seen previously, factors. Here, we have used the NFκB target genes that have been they also exhibit heterogeneous gene-expression patterns and are experimentally validated in humans and mouse (31) and serum- differentially down-regulated in the cells of different geometries inducible genes where MRTF–SRF directly bind and are sensi- upon TNFα stimulation (SI Appendix,Fig.S3M and Table S4). tive to the actin contractility (LatrunculinB and/or CytochalasinD sensitivity) (32). The expression profiles of the NFκBandMKL- Global Gene-Expression Profile Indicates the Presence of a Geometry- dependent SRF target genes obtained from other methods (32–34) Dependent Transcription Response to TNFα. The observed dependence aresummarizedinSIAppendix,SI Appendix,Fig.S3F–I (and SI of the gene-expression patterns of the NFκB and MKL-dependent Appendix,TablesS2andS3). A heatmap representing the relative SRF target genes in response to TNFα on cell geometry prompted gene-expression patterns of NFκB target genes (Fig. 3D) shows that us to explore the fate of the global transcription response under even before the stimulation of the cells with TNFα, CI cells have these conditions. As reported earlier (3), the gene-expression overall higher expression of these genes compared with RA cells. profiles were found to be very different for cells in the two ge- Following TNFα treatment, the expression of NFκB target genes ometries before treatment, and TNFα stimulation led to a differ- increased in both RA and CI cells. The overall trend of all these ential expression of 63 genes in RA and 94 genes in CI (SI genes reveals (Fig. 3E) that the transcription profile of the target Appendix,Fig.S3E). The genes that uniquely change in CI and RA genes follows the trend of the nuclear compartmentalization of the cells have been characterized and summarized in SI Appendix,Fig. relevant transcription factor. However, as indicated by the heat- S4 A–D and tabulated in SI Appendix,TableS5. To pick out ex- map, there is heterogeneity among the target genes (quantified in pression patterns that differ in the two geometries before and after SI Appendix,Fig.S3K, a). Whereas the global expression pattern TNFα stimulation, the ratio of expression in CI and RA cells be- correlatedwiththehighestrankfrequency in these samples, there fore and after treatment were calculated for all genes and those are still a considerable number of genes that do not follow the that have more than 30% difference (ratio ≤0.7 or ≥1.3) were global trend. Importantly, some of the genes that have their ex- identified as being differentially expressed (Fig. 4A). We found pression increased as a result of TNFα stimulation vary between three kinds of differential expression patterns in the system: genes RA and CI cells (SI Appendix,Fig.S3L and Table S4).This indi- that were similar in both the geometries before treatment and

6of10 | www.pnas.org/cgi/doi/10.1073/pnas.1618007114 Mitra et al. Downloaded by guest on September 27, 2021 A PNAS PLUS Type I Type II Type III /

TNF/ TNF B Before treatment After TNFα treatment TNF SCIENCES APPLIED PHYSICAL

TNF

Fig. 4. Global gene-expression profile indicates the presence of a geometry-dependent transcription response. (A and B) The three types of expression patterns visualized by plotting the Z scores of the genes (which are differentially expressed, i.e., having an expression ratio ≤0.7 or ≥1.3) in RA vs. CI cells before and after TNFα stimulation. (Purple: type I, genes similar in both the geometries before treatment and became different after TNFα stimulation; green: type II, genes that were different in both the geometries before treatment and became similar after TNFα stimulation; and red: type III, genes that were different in both the geometries before and after TNFα stimulation.) The gray dots represent the rest of the genes in the microarray. Data represented from three biological replicates.

became different after TNFα stimulation (type I), genes that were and, within 30 min, there is a change in the expression levels of different in both the geometries before treatment and became apoptotic and proliferative genes (SI Appendix,Fig.S5A, 3). similarafterTNFα stimulation (type II), and genes that were dif- Therefore, we explored TNFα-induced cell proliferation in the ferent in both the geometries before and after TNFα stimulation cells of the two geometries. RA and CI fibroblasts with and without (type III). These behaviors were visualized by plotting the Z score TNFα treatment were allowed to grow in the presence of EdU (5- of the gene expression in one geometry against the other under ethynyl-2´-deoxyuridine), which can integrate with and act as a unstimulated and TNFα-stimulated conditions (Fig. 4F)andare measure for active replicating DNA occurring during S phase of tabulated in SI Appendix,TableS6. In addition, the change in ex- cell cycle (36), for 9 h and examined for EdU incorporation pression of each gene that falls into the aforementioned three types (representative images, Fig. 5A and SI Appendix, Fig. S5B). CI cells is visually represented in SI Appendix,Fig.S4E and F and the showed more EdU incorporation than RA cells in untreated expression patterns of some representative genes are summarized conditions. Also, TNFα-treated RA cells showed significantly in SI Appendix,Fig.S4G. Gene ontology enrichment analysis of higher EdU incorporation than untreated RA cells (Fig. 5B). these genes sets reveals that many of these gene are related to However, CI cells showed no observed difference in EdU in- proliferative and cytoskeletal functions (SI Appendix,Fig.S4H and corporation with or without TNFα. This suggested that TNFα Table S8). The differential expression patterns of NFκBtarget stimulation enhanced S-phase DNA replication in RA cells in genes are summarized in SI Appendix,Fig.S4I. Collectively, these comparison with the CI cells. We further explored the relevance of gene-expression patterns highlight the importance of the cell ge- these differences toward cell proliferation and monitored cells ometry in modulating the response to the same cytokine in the grown in these conditions for 72 h, by when the geometrically same cell type into diverse types of transcription outputs, which confined individual cells within the micropattern divide and grow could potentially fine-tune cell behavior. into colonies (representative images, Fig. 5C and SI Appendix,Fig. S5C). After 72 h, cell proliferation was assayed by measuring the Geometry of the Cell Influences Proliferation in Response to TNFα. DAPI intensity per colony, which correlates with the number of Geometry-dependent significant differences in transcription out- nuclei per colony, for all surviving cells in the culture dish (rep- puts of cells induced by 30-min stimulation with TNFα led us to resentative images, Fig. 5D). Cells grown on rectangular micro- explore the subsequent long-term functional implication in terms patterns for 72 h showed higher cell numbers per colony compared of cell behavior. TNFα is known to regulate cell proliferation and with those grown on circular micropatterns. However, cells that apoptotic genes via NFκB and AP1 transcription regulators (35) were grown in the presence of TNFα on rectangular micropatterns

Mitra et al. PNAS Early Edition | 7of10 Downloaded by guest on September 27, 2021 ABUtr, EdU 9hrs TNF, EdU 9hrs *** 9hrs 1500000 *** F

1000000 Polarized Isotropic NS

500000

EdU intensity per nuclei 0

+ TNFα + TNFα Utr TNF Utr TNF

DAPI / EdU DAPI / EdU

Utr 72 hrs TNF 72 hrs CD Transcriptional Outputs 3hrs Response 1 Response 2

72 hrs Cell behaviour Cell behavioral Cell behavioral E Response 1 Response 2 72 hrs Utr *** TNF 300000 Utr F-actin Depolymerized MKL NFκB dimer TNF 200000 filament actin

100000 Cytoplasmic RNA polymerase II Target DNA IκB levels

DAPI intensity per colony DAPI 0 Utr TNF Utr TNF

Normalized frequencyNormalized DAPI n=788 n=1939 n=1120 n=2153

DAPI intensity per colony

Fig. 5. Geometry of the cell influences proliferation in response to TNFα.(A) Representative images of RA and isotropic NIH 3T3 cells, untreated or treated with TNFα (20 ng/mL), grown for 9 h in the presence of EdU, stained for DAPI and EdU (with Click-iT EdU Alexa Fluor 555 Imaging Kit). (Scale bar, 200 μm.) (B)Box-plots of fluorescence intensities of EdU per nuclei of RA and CI cells fixed and stained as in A. Data are presented as mean ± SD; box-plot represent 25–75 percentile; the small square within each box indicates mean; line indicates median ***P < 0.0001; N.S., not significant; Student’s t test; n = 738–1,666 nuclei in each conditions. (C) Representative image showing the nuclei of cells of the two geometries at 3 h and those of colonies formed after 72 h. (D) Representative images of colonies arisen from RA and CI NIH 3T3 cells, untreated or treated with TNFα (20 ng/mL), grown for 72 h stained for DAPI. (E) Probability distributions curves from his- togram showing fluorescence intensities of DAPI per colony arisen from RA and CI NIH 3T3 cells treated as in D and grown for 72 h. (Inset) Box-plots of fluorescence intensities of DAPI per colony of RA and CI cells grown, fixed and stained as in D. Data are presented as mean ± SD; box-plot represent 25–75 percentile; the small square within each box indicates mean; line indicates median ***P < 0.0001; Student’s t test; n = number of surviving colonies after 72 h in each condition is indicated. (F) Schematic showing the effect of cell geometry on the TNFα-induced cellular response.

showed an increase in the DAPI intensity per colony, whereas clear translocation of various transcription factors, one of which is those grown on circular micropatterns showed no increase com- NFκB. TNFα-induced processing of IκB leads to shuttling of the pared with unstimulated condition (Fig. 5E). In this experimental NFκB dimers into the nucleus where they trigger the expression of setup, we only measure the colonies that have survived for 72 h. To specific target genes. The shape, or wider mechanical state, of a cell account for the cells that undergo apoptosis and colonies that in a tissue is governed by its interaction with the matrix and detach, we quantified the fraction of cells that survived and found neighboring cells. In many pathophysiological conditions, the ECM that survival depended on the micropatterns and not on TNFα stiffness and cell–ECM interactions are altered, leading to changes stimulation. In addition, we found that TNFα stimulation induced in cell shape and behavior. In healthy connective tissues, fibroblasts apoptosis in colonies on both rectangular and circular micro- have an elongated and polarized shape which is mimicked by the patterns (SI Appendix,Fig.S5D). These observations suggests that cells grown in rectangular fibronectin micropatterns. Our earlier cell-geometry–TNFα interplay could possibly lead to interesting studies have shown that cells in two extremely different geome- regulatory balances between cell proliferation and cell death. tries––polarized and isotropic––differ in terms of various cellular Collectively, these results indicate that the differences in gene ex- properties like cytoskeletal architecture, nuclear and chromatin pression induced by the cell-geometry–TNFα interplay can trans- dynamics (10, 37). It was not known how cells of different shapes late into distinct cellular behaviors. respond to biochemical signals in the tissue microenvironment. We therefore studied the response of fibroblasts of two distinct me- Discussion chanical states––RA (stretched) and CI (relaxed), at the single-cell Cells in tissues are exposed to a multitude of mechanical and level, to TNFα-mediated signaling. biochemical stimuli from the microenvironment. They need to In contrast to CI cells, RA cells have robust F-actin stress fibers sense and respond to these stimuli correctly to maintain normal cell and higher levels of pMLC. The Rho-family-GTPases play a role physiology. Various cytokines in the connective tissue microenvi- in the regulation of actin polymerization through the activation of ronment modulate the behavior of fibroblasts by regulating distinct ROCK, phosphorylation of MLC, and the deactivation of cofilin signaling pathways. TNFα is an inflammatory cytokine known to by activation of LIM kinase-2 (23–26). We observed higher levels regulate cell proliferation, apoptosis, and is dysregulated in various of phospho-LIM kinase-2 and phospho-cofilin-Serine3 in RA forms of cancer. TNFα functions by regulating the activity or nu- compared with CI cells, which was indicative of increased LIM

8of10 | www.pnas.org/cgi/doi/10.1073/pnas.1618007114 Mitra et al. Downloaded by guest on September 27, 2021 kinase-2 activity, and reduced cofilin activity and thereby higher expression of MKL-dependent SRF target genes in both CI and PNAS PLUS actin polymerization. In addition, RA cells have relatively higher RA cells. This signifies that TNFα stimulation for 30 min can rewire amounts of cytoplasmic IκB and lower levels of nuclear NFκB the transcription machinery leading to changes in gene-expression (p65) compared with CI cells. Actin depolymerization alone in- programs. We also observed that the gene expression of αSMA, a duced by CytochalasinD is sufficient to induce IκB degradation target of MKL, changed in accordance with the differential loading and p65 nuclear translocation. These observations indicate that of MKL onto active transcription sites in the cells of the two ge- even in the absence of TNFα, the mechanical state of the cell, by ometries. Additionally, we also showed that the expression of modulating actin polymerization, can regulate the processing of αSMA is dependent on the nuclear exit of MKL following TNFα IκB and thereby nuclear translocation of NFκB (p65), suggesting a stimulation, in addition to the local inhibitory effect of p65 on possibility that TNFα-mediated signaling may involve similar MAL as reported earlier (42). mechanisms. This is corroborated by our observation that TNFα These observations prompted us to look at the global changes in induces a higher degree of F-actin depolymerization and phos- the transcription profiles of cells in the two geometries upon phorylation and degradation of IκB in the RA cells. Furthermore, TNFα stimulation. Although the global gene-expression trends of TNFα-induced p65 nuclear translocation significantly increased in the target genes of the transcription factors correlate with the the presence of actin depolymerization drugs. More importantly, enrichment of these factors in the active transcription sites in the the expression of a constitutively active Rho was associated with nucleus, we do observe that there are certain genes that do not compromised IκB processing and p65 nuclear translocation. follow this trend. These observations prompted us to look at the Collectively, our observations show TNFα-mediated F-actin global changes in the transcription profiles of cells in the two depolymerization via Rho inhibition is associated with IκB geometries upon TNFα stimulation. We find that the genes that degradation leading to NFκB (p65) nuclear translocation. are differentially expressed following TNFα stimulation are dif- TNFα-mediated actin remodeling and regulation of Rho ferent in the two geometries. Additionally, we show that there are GTPases has been reported earlier in the literature (13, 31, 38). In different classes of genes that respond to TNFα in a geometry- κ addition, the link between actin remodeling and NF B activity has specific manner (Fig. 4), suggesting that the transcriptional control been suggested earlier (12) However, a systematic analysis of of these genes is different in the two geometries. Transcription α TNF -mediated F-actin depolymerization and the mechanism by activity has been shown to depend on interchromosome archi- κ which this structural change could lead to the activation of NF B tecture (43). Because the chromatin structure and dynamics are

pathway at single-cell levels was lacking in the field. In our study we different in cells of these two geometries (37, 44), we speculate SCIENCES describe a mechanism that shows how TNFα-induced depoly-

that the observed differences in the transcription is due to the APPLIED PHYSICAL merized F-actin structures (actin nodes) could possibly regulate the differential chromosome architecture in the cells of two mechan- κ biochemical processing of I B. In support of this mechanism, we ical states. Finally, we also demonstrated geometry-dependent observe that phosphorylated IκB is enriched on the actin nodes in α α response to TNF -induced cell proliferation, observed 3 d after CI cells. Following TNF treatment, the actin nodes generated by treatment, indicating that the differential transcription outputs are TNFα induced F-actin depolymerization in RA cells are also κ translated into distinct long-term cellular response. enriched in phosphorylated I B. This mechanism could also explain Thus, we demonstrate that, depending on the geometry of the why the inherent differences in the actin-polymerization states of cell, TNFα signaling can lead to the compartmentalization of the thesamecelltypeinthetwogeometries can create differences in κ κ transcription regulators NF B and MKL by modulating actin po- nuclear translocation of NF B and transcriptional response. lymerization. This in turn leads to the rewiring of the transcription Unstimulated CI cells have actin in relatively depolymerized form machinery by differential loading of the two transcription factors with more nodes which enable them to process the signal in an onto the RNA polymerase II enzyme, resulting in the transcription early, fast, and efficient manner. In contrast, in RA cells, actin of their target genes (Fig. 5F). Collectively, our results highlight the nodes that are associated with IκB phosphorylation need to be importance of the intrinsic mechanical properties of a cell in de- generated by TNFα-mediated actin depolymerization, leading to a termining its response to a biochemical signal. Because cells ex- delay in the process. perience distinct forces within the tissue microenvironment, their Actin polymerization is known to regulate the nuclear localiza- geometry-dependent differential response to various cytokines are tion of the SRF cofactor, MKL (39), an effector of TGFβ, a cy- important in the maintenance of their cellular homeostasis. More tokine which has been shown to be antagonistic to TNFα (40, 41). importantly, this indicates that the cellular response to biochemical Cell-geometry-dependent differential actin organization leads to signals is regulated by cell-geometric constraints. Alterations to this differential MKL nuclear localization, with the RA cells having a cellular homeostasis may result in physiological abnormalities at significant amount of nuclear MKL, whereas CI cells have largely cytoplasmic MKL. As a consequence of TNFα-induced F-actin single-cell level, leading to diseases like fibrosis and cancer. depolymerization, MKL is shuttled out of the nucleus, which is Materials and Methods prominent in the RA cells owing to their high F-actin content. α Thus, depending on the mechanical state of the cell, TNFα induces Details of micropatterning, culturing of NIH 3T3 on the micropatterns, TNF “ ” “ ” treatment, immunostaining, confocal imaging, FRAP assays, cell proliferation, nuclear shuttling in of p65 and shuttling out of MKL, thereby DNA replication, and cell death assays are included in SI Appendix, Materials creating a distinct compartmentalization of MKL and p65. Tran- and Methods. Details of the various kinds of image analysis used in this study scription factors act by binding to specific sequences of their target (i.e., estimation of the nuclear fraction of the transcription regulators, coloc- genes to induce transcription. We show that, in addition to creating alization studies, segmentation of actin nodes and filaments) have been de- compartmentalization of these transcription regulators between the scribed in SI Appendix, Materials and Methods. Additionally, the details of the nucleus and the cytoplasm, the TNFα–cell-geometry interplay also methodologies used in the gene-expression studies, microarray sample prepa- leads to differential loading of these factors to the active RNA ration, and analysis have been provided in SI Appendix Materials and Methods. polymerase II transcription machinery. Consistent with the above, we observed that the NFκB target genes have overall higher ex- ACKNOWLEDGMENTS. We thank Caroline Uhler for useful discussions. We also thank Steven Wolf for critical reading of the manuscript. We thank the pression in CI whereas the MKL-dependent SRF genes have Ministry of Education Tier-3 program (2012-T3-1-001) and Institute of Molec- higher expression in RA. Following TNFα stimulation, we observe ular Oncology, Italian Foundation for Cancer Research and Mechanobiology an overall increase in the expression of NFĸB and a decrease in the Institute Joint-Research Laboratory for funding.

Mitra et al. PNAS Early Edition | 9of10 Downloaded by guest on September 27, 2021 1. Wang N, Tytell JD, Ingber DE (2009) Mechanotransduction at a distance: Mechanically 23. Sit ST, Manser E (2011) Rho GTPases and their role in organizing the actin cytoskel- coupling the extracellular matrix with the nucleus. Nat Rev Mol Cell Biol 10:75–82. eton. J Cell Sci 124:679–683. 2. Humphrey JD, Dufresne ER, Schwartz MA (2014) Mechanotransduction and extra- 24. Sumi T, Matsumoto K, Nakamura T (2001) Specific activation of LIM kinase 2 via phos- cellular matrix homeostasis. Nat Rev Mol Cell Biol 15:802–812. phorylation of threonine 505 by ROCK, a Rho-dependent protein kinase. JBiolChem 3. Jain N, Iyer KV, Kumar A, Shivashankar GV (2013) Cell geometric constraints induce 276:670–676. modular gene-expression patterns via redistribution of HDAC3 regulated by acto- 25. Scott RW, Olson MF (2007) LIM kinases: Function, regulation and association with myosin contractility. Proc Natl Acad Sci USA 110:11349–11354. human disease. J Mol Med (Berl) 85:555–568. 4. Dupont S, et al. (2011) Role of YAP/TAZ in mechanotransduction. Nature 474:179–183. 26. Mizuno K (2013) Signaling mechanisms and functional roles of cofilin phosphoryla- 5. Iyer KV, Pulford S, Mogilner A, Shivashankar GV (2012) Mechanical activation of cells in- tion and dephosphorylation. Cell Signal 25:457–469. duces chromatin remodeling preceding MKL nuclear transport. Biophys J 103:1416–1428. 27. Luo W, et al. (2013) Analysis of the local organization and dynamics of cellular actin 6. Engler AJ, Sen S, Sweeney HL, Discher DE (2006) Matrix elasticity directs stem cell networks. J Cell Biol 202:1057–1073. lineage specification. Cell 126:677–689. 28. Miralles F, Posern G, Zaromytidou AI, Treisman R (2003) Actin dynamics control SRF – 7. Kilian KA, Bugarija B, Lahn BT, Mrksich M (2010) Geometric cues for directing the activity by regulation of its coactivator MAL. Cell 113:329 342. differentiation of mesenchymal stem cells. Proc Natl Acad Sci USA 107:4872–4877. 29. Cisse II, et al. (2013) Real-time dynamics of RNA polymerase II clustering in live human – 8. Desprat N, Supatto W, Pouille PA, Beaurepaire E, Farge E (2008) Tissue deformation cells. Science 341:664 667. modulates twist expression to determine anterior midgut differentiation in Dro- 30. Crider BJ, Risinger GM, Jr, Haaksma CJ, Howard EW, Tomasek JJ (2011) Myocardin- β sophila embryos. Dev Cell 15:470–477. related transcription factors A and B are key regulators of TGF- 1-induced fibroblast – 9. Butcher DT, Alliston T, Weaver VM (2009) A tense situation: Forcing tumour pro- to myofibroblast differentiation. J Invest Dermatol 131:2378 2385. 31. Wójciak-Stothard B, Entwistle A, Garg R, Ridley AJ (1998) Regulation of TNF-alpha-induced gression. Nat Rev Cancer 9:108–122. reorganization of the actin cytoskeleton and cell-cell junctions by Rho, Rac, and Cdc42 in 10. Li Q, Kumar A, Makhija E, Shivashankar GV (2014) The regulation of dynamic me- human endothelial cells. J Cell Physiol 176:150–165. chanical coupling between actin cytoskeleton and nucleus by matrix geometry. 32. Esnault C, et al. (2014) Rho-actin signaling to the MRTF coactivators dominates the Biomaterials 35:961–969. immediate transcriptional response to serum in fibroblasts. Genes Dev 28:943–958. 11. Wajant H, Pfizenmaier K, Scheurich P (2003) Tumor necrosis factor signaling. Cell 33. Selvaraj A, Prywes R (2004) Expression profiling of serum inducible genes identifies a Death Differ 10:45–65. subset of SRF target genes that are MKL dependent. BMC Mol Biol 5:13. 12. Németh ZH, et al. (2004) Disruption of the actin cytoskeleton results in nuclear factor- 34. Gosselin K, Touzet H, Abbadie C (2004) Rel/NF-kappaB target genes. Available at kappaB activation and inflammatory mediator production in cultured human in- bioinfo.lifl.fr/NF-KB/. Accessed January 19, 2017. testinal epithelial cells. J Cell Physiol 200:71–81. 35. Baud V, Karin M (2001) Signal transduction by tumor necrosis factor and its relatives. 13. Goldblum SE, Ding X, Campbell-Washington J (1993) TNF-alpha induces endothelial Trends Cell Biol 11:372–377. cell F-actin depolymerization, new actin synthesis, and barrier dysfunction. Am J 36. Salic A, Mitchison TJ (2008) A chemical method for fast and sensitive detection of DNA – Physiol 264:C894 C905. synthesis in vivo. Proc Natl Acad Sci USA 105:2415–2420. 14. Barkett M, Gilmore TD (1999) Control of apoptosis by Rel/NF-kappaB transcription 37. Makhija E, Jokhun DS, Shivashankar GV (2016) Nuclear deformability and telomere dy- – factors. Oncogene 18:6910 6924. namics are regulated by cell geometric constraints. Proc Natl Acad Sci USA 113:E32–E40. 15. Karin M, Lin A (2002) NF-kappaB at the crossroads of life and death. Nat Immunol 3: 38. Mong PY, Petrulio C, Kaufman HL, Wang Q (2008) Activation of Rho kinase by TNF- – 221 227. alpha is required for JNK activation in human pulmonary microvascular endothelial 16. Micheau O, Tschopp J (2003) Induction of TNF receptor I-mediated apoptosis via two cells. J Immunol 180:550–558. – sequential signaling complexes. Cell 114:181 190. 39. Olson EN, Nordheim A (2010) Linking actin dynamics and gene transcription to drive 17. Aggarwal BB (2003) Signalling pathways of the TNF superfamily: A double-edged cellular motile functions. Nat Rev Mol Cell Biol 11:353–365. sword. Nat Rev Immunol 3:745–756. 40. Verrecchia F, Mauviel A (2004) TGF-beta and TNF-alpha: Antagonistic cytokines con- 18. Li YP, Schwartz RJ (2001) TNF-alpha regulates early differentiation of C2C12 myo- trolling type I collagen gene expression. Cell Signal 16:873–880. blasts in an autocrine fashion. FASEB J 15:1413–1415. 41. Distler JH, Schett G, Gay S, Distler O (2008) The controversial role of tumor necrosis 19. Sero JE, et al. (2015) Cell shape and the microenvironment regulate nuclear trans- factor alpha in fibrotic diseases. Arthritis Rheum 58:2228–2235. location of NF-κB in breast epithelial and tumor cells. Mol Syst Biol 11:790. 42. Tang RH, et al. (2008) Myocardin inhibits cellular proliferation by inhibiting NF- 20. Iskratsch T, Wolfenson H, Sheetz MP (2014) Appreciating force and shape—the rise of kappaB(p65)-dependent cell cycle progression. Proc Natl Acad Sci USA 105:3362–3367. mechanotransduction in cell biology. Nat Rev Mol Cell Biol 15:825–833. 43. Maharana S, et al. (2016) Chromosome intermingling-the physical basis of chromo- 21. Shivashankar GV (2010) Nuclear mechanics and genome regulation. Methods Cell Biol some organization in differentiated cells. Nucleic Acids Res 44:5148–5160. 98:xiii. 44. Toh KC, Ramdas NM, Shivashankar GV (2015) Actin cytoskeleton differentially alters 22. Oeckinghaus A, Ghosh S (2009) The NF-kappaB family of transcription factors and its the dynamics of lamin A, HP1alpha and H2B core histone proteins to remodel chro- regulation. Cold Spring Harb Perspect Biol 1:a000034. matin condensation state in living cells. Integr Biol 7:1309–1317.

10 of 10 | www.pnas.org/cgi/doi/10.1073/pnas.1618007114 Mitra et al. Downloaded by guest on September 27, 2021